An Intracardiac Soft Robotic Device for Augmentation of Blood Ejection

from the Failing Right Ventricle

MARKUS A. HORVATH,1,2,4 ISAAC WAMALA,2 ERIC RYTKIN,2 ELIZABETH DOYLE,3 CHRISTOPHER J. PAYNE,1,4

THOMAS THALHOFER,1 IGNACIO BERRA,2 ANNA SOLOVYEVA,2 MOSSAB SAEED,2 SARA HENDREN,3

ELLEN T. ROCHE,1,4 PEDRO J. DEL NIDO,2 CONOR J. WALSH,1,4 and NIKOLAY V. VASILYEV2

1Wyss Institute for Biologically Inspired Engineering at Harvard University, Boston, MA, USA; 2Boston Children’s Hospital,Harvard Medical School, Boston, MA, USA; 3Olin College of Engineering, Needham, MA, USA; and 4Harvard John A.

Paulson School of Engineering and Applied Sciences, Cambridge, MA, USA

(Received 11 January 2017; accepted 11 May 2017)

Associate Editor K. A. Athanasiou oversaw the review of this article.

Abstract—We introduce an implantable intracardiac softrobotic right ventricular ejection device (RVED) for dynamicapproximation of the right ventricular (RV) free wall and theinterventricular septum (IVS) in synchrony with the cardiaccycle to augment blood ejection in right heart failure (RHF).The RVED is designed for safe and effective intracardiacoperation and consists of an anchoring system deployedacross the IVS, an RV free wall anchor, and a pneumaticartificial muscle linear actuator that spans the RV chamberbetween the two anchors. Using a ventricular simulator and acustom controller, we characterized ventricular volumeejection, linear approximation against different loads andthe effect of varying device actuation periods on volumeejection. The RVED was then tested in vivo in adult pigs(n = 5). First, we successfully deployed the device into thebeating heart under 3D echocardiography guidance (n = 4).Next, we performed a feasibility study to evaluate thedevice’s ability to augment RV ejection in an experimentalmodel of RHF (n = 1). RVED actuation augmented RVejection during RHF; while further chronic animal studieswill provide details about the efficacy of this support device.These results demonstrate successful design and implemen-tation of the RVED and its deployment into the beatingheart. This soft robotic ejection device has potential to serveas a rapidly deployable system for mechanical circulatoryassistance in RHF.

Keywords—Right heart failure, Soft robotics, Mechanical

circulatory support.

INTRODUCTION

The right ventricle (RV) of the heart pumps venousblood collected from the body to the lungs for oxy-genation.17 The RV has a crescentic shape and isbordered by a thin free wall and a rightward convexinterventricular septum (IVS) (Fig. 1a). RV contrac-tion is a complex 3-dimensional motion that includeswave-like contraction of the RV free wall, approxi-mation of the RV free wall toward the IVS andthickening of the IVS, which results in blood ejectiontowards a low-resistance pulmonary vascular bed.4,9,17

Right heart failure (RHF) is a disease entity withdiverse etiology, where RV pumping function is se-verely depressed.10,11,17,27 In most cases of advancedRHF, the RV free wall and the IVS are less contractile,the RV is dilated and the septum is no longer convexshaped but rather flattened or even shifted toward theleft ventricle (LV) (Fig. 1b).6,10

RHF is characterized by high morbidity and mor-tality.15,18,27 The current treatment options include awide spectrum of approaches, from pharmacologicaltherapy in the early stages, to heart or heart–lungtransplantation in end stage of the disease.10,27 In ad-vanced RHF, mechanical circulatory support (MCS)has become a valuable therapeutic solution, when adonor organ is not available. The state of the artamong MCS devices is one-way flow pumps that divertblood returning from the body to pulmonary vascu-lature.1,12

We present an alternative approach where a softrobotic pneumatic artificial muscle (PAM) is deployedacross the failing RV free wall and IVS to augmenttheir approximation during systole, leading to ejection

Address correspondence to Nikolay V. Vasilyev, Boston Chil-

dren’s Hospital, Harvard Medical School, Boston, MA, USA.

Electronic mail: nikolay.vasilyev@childrens.harvard.edu

Markus A. Horvath and Isaac Wamala have contributed equally

to this work and are co-first authors.

Annals of Biomedical Engineering (� 2017)

DOI: 10.1007/s10439-017-1855-z

� 2017 Biomedical Engineering Society

of additional blood volume from the RV (Figs. 1c and1d). Thus, pulsatile flow augmentation is achievedacross the native cardiovascular vasculature withoutdiverting blood flow through an artificial circuit.

McKibben PAMs have been shown to have forcecontraction characteristics similar to muscle7,8 andhave demonstrated the ability to support the failingheart by closely mimicking the contractile motion ofthe heart.22,23 In contrast to previous designs thatprovide compressive biventricular support on the heartsurface, the proposed RVED operates inside the RV,specifically addressing the structural changes that areseen in RHF.20 A low profile device such as this, whichcan be easily deployed into the beating heart underechocardiography guidance, may be very well suitedfor the short-term support of a failing RV. Here wepresent the design and fabrication of the device, as wellas ex vivo and in vivo testing in a porcine animal model.

MATERIALS AND METHODS

Device Design and Fabrication

The RVED is comprised of four major components:a linear actuator, a septal anchoring system, an RVfree wall anchoring system and seal (Figs. 1c, 1d, and

1e), and a controller. We approached the device designby addressing each component separately. For eachcomponent, we defined the functional requirements,and then developed engineering solutions.

Linear Actuator

The linear actuator is a modified McKibben PAMthat consists of an inflatable inner bladder constrainedby a braided polyethylene terephthalate (PET) mesh.When the inner bladder of the actuator is inflated itsradial expansion is constrained by the outer mesh,which results in linear contraction (Fig. 1d).20 ThePAM is customized for intracardiac use by fabricatingit entirely from polymer materials to allow echocar-diography guided placement into the beating heart andby the addition of an outer bladder that isolates thedevice from the blood (Fig. 1e). Additionally, applyingvacuum to the outer bladder during RV diastole, whilethe internal bladder is deflated, facilitates PAMextension and minimizes the device volume thusallowing RV diastolic filling. Both bladders are actu-ated independently through a double lumen airline(Fig. 1e) attached to a custom control system.

Previously suggested PAMs for cardiac supportprovided up to 24% linear contraction at a total lengthof 140 mm.19 Roche et al. described the use of non-compliant thermoplastic polyurethane (TPU) bladdersto improve reliability while providing up to 34% linearcontraction in a 120 mm long PAM for cardiacassist.21 To meet the specifications of the requiredintracardiac use, the PAM was customized to a 10 mmmaximum width and 40 mm length which was thelower limit of a dilated human RV.5,14 To match thetypical free wall to IVS approximation distance seenduring RV ejection, the required linear contraction ofthe PAM was determined to be 25%.5,9,14 This con-traction works against the elevated hemodynamicpressure in the RV and the afterload in the pulmonaryartery. In contrast to previously suggested PAMs forcardiac support that are actuated at pressures close toatmospheric pressure in the chest cavity, the RVEDoperates inside the heart. During RHF blood pressurein the RV and pulmonary artery can reach 70 mmHg.This pressure would diminish the relative inflatingpressure of the RVED internal bladder compromisingPAM contractibility. Accordingly, the RVED wasrequired to exert contraction forces higher than thatrequired for external cardiac support devices (50 N).

The inner and outer bladder (0.3 and 0.25 mm thickTPU respectively) were fabricated as previouslydescribed.21 To avoid air leaks the inner bladder washeat sealed to its airline (TPU, Eldon James Corpo-ration) with a custom milled aluminum mold heated to180 �C for 40 s. Injection molding techniques were

FIGURE 1. (a) Illustration of the cross section of a healthyheart, in comparison with (b) a heart with right heart failurecharacterized by a dilated RV, deformed LV and ventricularseptal right-to-left shift. (c) Schematic illustration of the rightventricular ejection device implanted in a heart with rightheart failure. (d) The actuation of the pneumatic artificialmuscle results in linear approximation (black arrows) of theventricular septum and the RV free wall. (e) Schematic illus-tration of the linear actuator cross section depicting thecomponents.

HORVATH et al.

used to fabricate the proximal and distal part of theRVED device. A two-step process was employed using3-D printed molds (Vero material, Objet Connex 500)as shown in Figs. 2. Figures 2a and 2b show theinjection mold process for the distal end (9.8 mmdiameter, 4 mm length molded part). Figure 2c showsthe injection molding process for the proximal end(9.8 mm diameter, 40 mm length molded part). Toensure homogenous polymer formation each injectionmold had an overflow tract, and the resin was cured ateach step under 400 kPa pressure for twelve hours. Thefinal device is shown in Fig. 2d.

Septal and Free Wall Anchoring Systems

For the septal anchoring system, we defined thefollowing requirements: linear deployment across theIVS without creating an interventricular shunt; abilityto withstand the force generated by the linear actuator;ultrasound compatibility; ability to involve an ade-quate area of the septum in its approximation towardthe RV free wall.

The anchor consists of collapsible wings that wereradially bound on a stainless-steel base around athreaded core pin, such that the wings can only unfoldup to 90� in order to avoid folding backwards duringthe device actuation (Figs. 3a and 3b) To provide en-hanced ultrasound compatibility and withstandhigh dynamic forces, the wings were fabricated from ahigh modulus polymer (Polyether ether ketone—

PEEK—7 mm 9 1.5 mm 9 2 mm). A superelasticnickel-titanium (Nitinol) wire ring was attachedthrough the free ends of the wings to form a circularspring system and enable self-expanding deployment.A second retrievable Nitinol release wire was attachedto the ring allowing controlled device repositioning(see Supplementary Video). Once the wings are de-ployed in the appropriate position and laid flat on theleft side of the IVS, a T shaped cap is secured on thecenter pin of the anchor from the RV side. The IVS isthen sandwiched between the open wings on the leftside and the cap on the right side. Enclosing the IVS inthis manner seals the access area in the IVS and pre-vents intraventricular shunting.

To secure the actuator to the RV free wall, and toprevent blood loss during device actuation, we used atubular RV free wall anchor (Fig. 3b). The free wallwas sandwiched between the inner lip of the anchorand the external purse string suture. Additionally, anO-ring positioned between the anchor shaft and the airsupply lines sealed the device from inside preventingblood loss. A radially positioned stud screw fixed theRV free wall anchor to the air supply lines. The entirecustom design of the anchor was 3D printed in Veromaterial (Objet Connex 500).

Actuation Controller

A custom software interface (Fig. 3c) was developed(LabVIEW, National Instruments) using an X-Series

FIGURE 2. Schematic of the device fabrication process: (a) device tip: outer bladder and mesh are collapsed over a nylon screwand fixed by a TPU locking ring before positioning in the injection mold; (b) cross section during injection molding; (c) devicebase: insertion of inner bladder together with supply tubes and polymer casting; (d) photo of finalized contraction device.

An Intracardiac Soft Robotic Device for Augmentation

Data Acquisition Box (DAQ) (National Instruments).Pressurization of the internal bladder as well as theapplication of negative relative pressure to the outerbladder were independently programmable by settingdesired pressure levels, actuation periods and frequen-cies. Utilizing two electro-pneumatic regulators (ITVseries, SMC) coupled to separate solenoid valves (VQseries high-flow 3 way, SMC) the control system deliv-ered compressed air and vacuum from standard wallsources to the actuator. Synchronization to the nativecardiac cycle was established employing a pacemaker(5342, Medtronic Inc.) which simultaneously sent anelectrical signal to the RV and to the controller.

Component Optimization and Force Characterization

A single filament braided PET mesh with maximaldiameter of 26 mm (Flexo PTO 0.25, Techflex Inc.)and 40 mm in length was selected. Three mesh/bladdercombinations were manufactured using the TPU innerbladders 20, 30 and 40 mm in length placed inside the

mesh. The bladders were detached from the distal endof the mesh.

Three specimens of each combination were actuateddynamically for 250 ms contraction period at 1.25 Hz(simulating physiological ranges of 75 beats per min-ute) while input pressures were systematicallyincreased from 20 to 240 kPa. Actuation was videorecorded at 250fps. Single frames at maximal con-traction and relaxation were extracted, and linearcontraction was calculated using image-processingsoftware (ImageJ, National Institutes of Health).

The highest performing mesh/bladder combinationfrom this experiment was then tested (n = 3) to mea-sure the force/contraction behavior with a mechanicaltensile tester (Instron 5566, Instron, Norwood, USA).Specimens were secured in the fixation clamps of thetester and peak forces were recorded at contractionlevels from 0 to 35% under similar dynamic actuationas in the contraction study at 1.25 Hz. The inputpressure was systematically increased from 34 to275 kPa.

FIGURE 3. (a) Illustration of the delivery of the collapsed septal anchor into the LV in the beating heart and (b) of completeejection device positioned in the RV. (c) Custom software interface to control device actuation.

HORVATH et al.

To quantify extension forces of the external bladder,the final actuator design was tested in a tensile tester.After device fixation in the testing machine a vacuumpressure of 80 mmHg was dynamically applied to theouter bladder (Fig. 1e) and the resulting force wasrecorded.

Performance Characterization in Ventricular Simulator

To characterize device functionality under theenvisaged hemodynamic conditions, we designed aventricle simulator. The simulator consisted of twoopposing silicone membranes supported by an acrylicframe (Fig. 4a). The membrane material (Ecoflex00-30, Smooth-on Inc.) was chosen to imitate the stiffnessof heart tissue23 and the wall thickness was matched tothe respective thickness of the RV free wall (5 mm) andthe IVS (15 mm) as determined from measurements inporcine hearts (n = 3). The total surface area of thesilicone walls was designed to match the total surfacearea in a typical RV (130 cm2) as calculated from CTscan data of a human heart and examinations of por-cine hearts (n = 3). We used water in lieu of blood andmeasured the volumes ejected via a 10 mm wide cali-brated outflow tube. Simulator chamber pressure wasmonitored in real-time (ArgoTrans model 2 Neonatalmacro-drip transducer, Argon medical; Surgivet,Smith Medical). Afterloads were varied from 10 to70 mmHg by changing the height of the water columnin the outflow tube. Since there was no valve betweenthe outflow tube and the chamber, afterload pressurewas equal to chamber pressure prior to actuation. Thedevice was actuated at 1.25 Hz (simulating physiolog-ical frequencies of 75 beats per minute) and inputpressures were systematically increased from 20 to275 kPa. Pressurization periods ranged from 100 to300 ms to assess the effect of actuation period onejected volumes. The experiment was video recorded,and single frames were extracted for analysis. Ejectionvolume and chamber pressure were measured directly.Linear contraction of the actuator was calculated(ImageJ software) employing a laser engraved mil-limeter scale in the acrylic frame for calibration(Fig. 4b).

Prior to in vivo device testing we conducted adurability study of the RVED in the ventricular sim-ulator for 6 h (n = 5). A duration of 6 h was chosen toassure that the device would be suitable for safe use inan intracardiac position during the in vivo evaluation.The simulator was filled with 0.9% saline solution(temp = 36.5 �C), and the device was actuated for250 ms periods at 1.25 Hz and maximal pressures of275 kPa. Visual inspection every 30 min was con-ducted to detect bubble formation in the water bath orany water in the airlines. At the end of each experi-

ment, the actuators were disassembled and inspectedfor wear.

In Vivo Evaluation

To assess device placement and anchoring underechocardiography guidance we used female Yorkshirepigs (n = 4). We then evaluated feasibility of deviceperformance in a porcine RHF model (n = 1). Theanimals received humane care in accordance with the1996 Guide for the Care and Use of Laboratory Ani-mals recommended by the US National Institute ofHealth. The experimental protocol was approved bythe Boston Children’s Hospital Institutional AnimalCare and Use Committee.

The procedures were performed under generalanesthesia, as previously described.25 The heart wasapproached via midline sternotomy and flow probeswere placed around the main pulmonary artery (PA)and the ascending aorta (Transonic PS series). RV, PAand systemic pressures were measured directly in real-time (Argo Trans probes).

Echocardiography Guided Implantation into BeatingHeart

The feasibility of RVED implantation under 2Dand 3D echocardiography guidance on the beatingheart was evaluated in four pigs. The target area forthe RVED placement was identified using echocar-diography guidance. For the septal anchor placement,we chose the mid-muscular ventricular septum. Bypositioning the device directly inferior to the septalpapillary muscle, the chance of affecting the atrioven-tricular nodal branches of the right coronary artery orthe first septal perforator branches of the left anteriordescending coronary artery is minimized.13 Addition-ally, device placement in this area avoids the conduc-tion system, the branches of the bundle of His,3 thusminimizing the risks associated with this placement.For the RV free wall anchor, the area just anterior tothe mid portion of the anterior papillary muscle wasselected.

After heparin administration (300 U/kg IV, acti-vated clotted time above 250 s), a purse-string suturewas placed on the RV free wall, and a marked needlewas introduced into the RV and then advanced intothe LV through the IVS. A guide wire was introducedinto the LV through the needle and the needle was thenremoved. The access area in the free wall and septumwas enlarged by a series of dilators over the wire. TheLV part of the septal anchor in the folded configura-tion, attached to a stainless-steel pusher rod (Fig. 3a),was delivered via a valved 20Fr introducer sheath(Cook medical) and opened (Fig. 3b) on the LV side of

An Intracardiac Soft Robotic Device for Augmentation

the IVS such that the opened wings lay flat against theseptum. Echocardiography guidance was used to avoiddamaging mitral valve chordae apparatus during an-chor placement, and retractable control of wing fold-ing and unfolding facilitated precise positioning. Next,a custom made 12 mm introducer sheath with the RVpart of the septal anchor was delivered into the RV viathe same access over the pusher, and the RV part of theseptal anchor was screwed onto the threading of theLV part, which completed the septal anchor place-ment. Then, the pusher was replaced with the actuator(Fig. 3b), the sheath was withdrawn, and the RV freewall anchor was placed into the position and securedwithin the purse string suture. The device was thenconnected to the air and vacuum source and operatedusing the controller.

Functional Evaluation in the Model of End-Stage RHFResulted from Volume Overload

In order to determine performance of the RVED inconditions of end-stage RHF resulting from volumeoverload, in vivo evaluation under cardiopulmonarybypass (CPB) was performed in one animal. Use ofCPB was necessary to safely simulate a status of RVvolume overload while leaving the systemic circulationintact. General anesthesia was established as describedabove. A midline sternotomy was performed, and CPBwas established such that venous return was collectedfrom the superior vena cava and the inferior vena cava,and arterial inflow was returned to the right femoralartery. During CPB fentanyl (50–200 m/kg) andmidazolam (0.2 mg/kg) were administered for anes-thesia. The animals were cooled to 32 �C, at whichpoint the ascending aorta was cross-clamped and theheart was arrested by anterograde delivery of coldcardioplegia solution (del Nido solution). The rightatrial appendage was opened and the RVED was im-planted under direct vision after which the right atri-otomy was closed, the heart was de-aired, the animalwas rewarmed and the aortic cross-clamp was removed

to restart heart contraction. The animal was main-tained on 80% CPB flow. By controlling the amount ofthe venous return into the right side of the heart (CVPup to 16 mmHg), we could generate RV volumeoverload in controllable fashion causing RHF. Thehemodynamic effect of the RVED actuation wasassessed by measuring PA and aortic flow, PA, RV andsystemic pressures. The device was actuated in syn-chrony with heart contractions that were triggered byepicardial pacing; the same pacing signal was sent tothe heart and to the controller for triggering the device.During the device actuation, the data was recorded for10 min. Data were averaged over 10 cardiac cycles foranalysis.

RESULTS

Actuator Shortening and Contraction Forces

Contraction behaviors at varying inner bladder tomesh lengths are summarized in Fig. 5a Reducingbladder to mesh length ratio led to improved linearcontraction at lower actuation pressures. PAM with abladder length of 50% compared to total mesh lengthreached 25% contraction at 60 kPa, and the maximalcontraction was 31% at 240 kPa. Actuators with 75%bladder length to mesh length ratio reached 25%contraction at 120 kPa, 28% contraction at 140 kPaand improved their maximal linear contraction to 31%at 241 kPa. Specimens with equal bladder and meshlength reached lower contractions at all actuationpressures with the maximal contraction of 27% at276 kPa (Fig. 5a). We observed material failure ofspecimens with 50% length ratio at input pressures of276 kPa but not with 75% or 100% ratios. For thefinal design we thus chose a bladder to mesh ratio of75%.

Generated force to contraction behavior is sum-marized in Fig. 5b. A peak contraction force of 140 Nwas generated by the device at 276 kPa. At 25% con-

FIGURE 4. (a) In vitro testing of the RVED in a simplified RV simulator; (b) contracted device is approximating the simulatedseptum and free wall.

HORVATH et al.

traction, at least 10 N of force could be generated byany input pressure above 69 kPa. Diastolic extensionof the actuator resulted in 6 N extension force.

Device Performance in Ventricular Simulator

Device performance for various input parametersare summarized in Fig. 6. The device was capable ofejecting 21 mL of volume at 10 mmHg afterload whichdropped 61% to 8 mL at 70 mmHg afterload(Fig. 6a). Device actuation was able to generate pres-sure in the simulator chamber of 20–28 mmHg(Fig. 6b). Maximal device shortening decreased from30% contraction at an afterload of 10 mmHg to 26%at an afterload of 70 mmHg (213.5%) (Fig. 6c).Increasing the input pressure above 180 kPa lead tominor improvements in ejection volume and generatedpressure. At 1.25 Hz increasing the contraction periodfrom 100 to 250 ms augmented ejected volume by 30%while additional increase to 300 ms did not result inhigher ejected volume. This implied an optimal actu-ation period of 250 ms. All components of the device(balloon, mesh, airline and bonding materials) werefound to be intact upon examination after completionof the durability studies.

Echocardiography Guided Implantation into the BeatingHeart

Echocardiography provided clear visualization ofthe anatomic structures and the septal anchor duringdevice placement (Fig. 7a). The device was successfullyimplanted across the RV in all four animals (Figs. 7band 7c). The deployment was typically completed in20 min. No damage to the intracardiac structures and

no major blood losses were observed. The linearactuator was securely connected to the septal and theRV anchoring systems in all animals. There was noblood loss at the RV access throughout the deviceoperation. One animal had an episode of heart blockafter device placement, which required therapeutic useof a pacemaker. Postmortem evaluation of the heartsdemonstrated successful sandwiching of the septum bythe septal anchor in all cases (Fig. 7d) without grossdamage to the heart.

In all deployment experiments, normal hemody-namics were preserved. There was no improvement inmean PA flow (1.6 l/min/SD 0.79 vs. 1.6 l/min/SD0.93respectively), mean RV pressure (9.8 mmHg/SD4.9 vs.10.1 mmHg/SD0.4), and mean PA pressure(17.9 mmHg/SD3.1 vs. 18.1 mmHg/SD2.8 respec-tively), over native RV function upon device actuationin these hearts with normal size and function.

Functional Evaluation in the Model of End-Stage RHFResulted from Volume Overload

The device successfully augmented RV ejection(Fig. 8). Systolic RV pressure decreased from21 mmHg baseline to 3.5 mmHg (17%) during RHFand was recovered to 29 mmHg (138%) during RVEDassistance (Figs. 8a and 8b). Diastolic RV pressureincreased during RHF from 10 mmHg baseline to24 mmHg and was increased to 28 mmHg during thedevice actuation (Fig. 8b). Systolic PA pressuredecreased from 10 mmHg baseline to 1.6 mmHg(17%) during RHF and was recovered to 8.8 mmHg(89%) by the RVED (Figs. 8c and 8d). Right ventric-ular output (PA flow) dropped during the RHF from2.4 L/min at healthy baseline to non-pulsatile 0 L/min

FIGURE 5. (a) Contraction characterization at varying actuation pressure of PAM with varying inner bladder length at constantmesh length (40 mm). (b) Contraction vs. Force curve for final RVED (40 mm) at increasing actuation pressures. The targeted 20 Nthreshold line is indicated (black dotted line).

An Intracardiac Soft Robotic Device for Augmentation

during CPB (not registered by the flow probe). RVEDactuation clearly increased the PA flow to 0.2 L/min(8% of healthy baseline) (Figs. 8e and 8f).

DISCUSSION

We demonstrate an alternative approach of assist-ing the failing ventricle using a device that is designedto augment native ventricular blood ejection. Aug-mentation is achieved by approximating the ventricu-lar free wall and the IVS in synchrony with the cardiaccycle. This strategy is especially relevant to RHFpatients with IVS flattening or right-to-left septalshift.6,27 Currently there are no small RVAD that canbe rapidly deployed that also provide pulsatile flowaugmentation to the native RV contraction. Theadvantages of pulsatile flow have been greatly studiedfor LVADs. Pulsatile VADs have been more or lessdisfavored because of their large size and propensityfor thrombi formation, especially at the one-way flowvalves. The proposed RVED is low profile and easily

deployed on the beating heart using echocardiographyguidance, the device achieves pulsatile flow augmen-tation while minimizing surface area of artificialmaterial in contact with blood. The maximal surfacearea of the device at the fully actuated state that in theblood flow path is 2500 mm.2

The concept of using soft robotic actuators mim-icking contractile function of ventricular walls hasbeen previously described by our group.19,22,23 ThePAM actuators are lightweight, have fast responsetimes, and are tunable to match physiologic require-ments of the heart.19,21 In contrast to previous designsthat provide biventricular support using an enclosingsleeve,22,23 the RVED is specifically conceived to ad-dress the structural changes that are seen in advancedright heart failure.6,10,20,27 In this study, by employinga combination of soft robotics and precision engi-neering we developed a device that is small enough tobe safely deployed inside the beating heart.

We have demonstrated that use of TPU internalbladders detached from the distal point of the meshand reduction of the bladder to mesh/ratio can lead to

FIGURE 6. (a) Ejected volume by device in ventricular simulator at increasing afterloads and actuation pressure. (b) Character-ization of generated pressure in the ventricular simulator upon device actuation under varying afterloads and increasing actuationpressure. (c) Linear contraction of device under varying afterloads and increasing actuation pressure. (d) Ejected volume atvarying actuation periods and increasing actuation pressure.

HORVATH et al.

improved actuator shortening but there is a thresholdbelow which actuator weakness is encountered. Wehave also demonstrated that the proposed design, atdynamic actuation of 1.25 Hz and input pressures ofabove 69 kPa, produced forces adequate to overcomethe expected physiologic hemodynamic resistance.McKibben PAMs intrinsically have self-limiting force–length curves similar to human muscle.7,8,16 Thismakes them one of the safest actuators to use for thispurpose, because the linear force applied decreaseswith contraction such that the maximal force appliedto the tissue will be as required to initiate contractionand not the total maximum force measured using atensile tester.

Using a ventricular simulator, we demonstrated thatit was feasible to obtain up to 21 mL of ejection vol-ume from the device while causing a linear approxi-mation of at least 26%. The device was operated at275 kPa for up to 6 h under physiological loads with-out undue fatigue further showing suitability tointracardiac use.

We have demonstrated the feasibility of deviceimplantation into the beating heart under echocar-diography guidance. In a model of end-stage RHFresulting in no ejection from the ventricle, the RVEDrestored ventricular pressures to baseline healthy val-ues and resulted in 8% augmentation of RV ejection,which may make a clinically significant difference inpatients with RHF. Van de Veerdonk et al. demon-

strated that in patients with PA hypertensionresponding to therapy with an improvement in pul-monary ventricular resistance, improvement in RVejection of 10% over the follow up period was signif-icantly associated with survival.26 Contraindication touse of this device would be biventricular failure, sincethe device augments only the RV contraction. Patientswith ventricular septal defects that would preclude safeimplantation of the anchoring system would not becandidates for this device. Lastly, patients with severetricuspid or pulmonary valve regurgitation may notbenefit from use of this device.

The main challenges in using soft robotic intracar-diac implants include the risk of air embolism in thecase of a leak from the pneumatic circuit, andrequirement of perfect synchronization of the deviceactuation with native heartbeat. Clinical experiencewith intra-aortic balloon pumps, that are placed insidethe aorta and actuated with inert gas, suggests that therisk of air leak can be significantly mitigated withengineering design.24 In our current prototype, wehave addressed the issue by customizing the PAMincorporating a second external protective bladder thatis not pressurized during the device actuation. Inaddition, vacuum is applied to the external bladderduring diastole, which may capture escaped air bubblesfrom the inner bladder serving as an additional pro-tective feature. An additional sensor system that de-tects pressure changes in the bladders in real time is

FIGURE 7. (a) 3D echocardiography image showing the septal anchor during in vivo delivery in the beating heart after thecontrolled opening of the anchor wings inside the LV. (b) Photo of the successfully deployed device into a beating heart. (c) 2Dechocardiography image showing implanted device inside RV and relation to the IVS. (d) Post mortem photo of the devicesuccessfully deployed into the beating heart displaying no visible damage to the septum.

An Intracardiac Soft Robotic Device for Augmentation

necessary. In this feasibility study, we synchronized thedevice actuation with the cardiac cycle by simultane-ously pacing both the RV and the device. Furtheroptimization of a synchronization closed loop algo-rithm using triggering off an EKG signal, RV pressurecurve and flow curve will improve synchronization.

While we describe the device design and demon-strate functional feasibility in one animal, further

animal studies are required to fully elucidate thephysiological impact of this approach. The RVED isfixed at distinct points on the RV free wall and at theseptum, and therefore the device contributes solely toRV free wall ventricular septum approximation. Innormal or thin walled RV, ballooning of the free wallareas not involved by the device hypothetically limitsthe effectiveness of the approach. Conversely,

FIGURE 8. RVED testing in RHF model resulting from volume overload. (a) Example of RV pressure showing periods of deviceactuation (A) and pause (P). (b) RV pressure in the healthy heart, the failing heart and the failing heart during device actuationconditions. (c) Example of PA pressure during heart failure showing periods of device actuation (A) and pause (P). (d) PA pressurecomparing the healthy heart, the failing heart and the failing heart during device actuation conditions. (e) Example of PA flowduring heart failure showing periods of device actuation (A) and pause (P). (f) PA flow comparing the healthy heart, the failing heartand the failing heart during device actuation conditions.

HORVATH et al.

implantation of the device into a hypertrophied RVwould improve augmentation in comparison with theresults we have observed in the normal ventricle. An-other strategy to improve ejection augmentation wouldbe the placement of additional actuator(s) to affect RVlongitudinal shortening. While no macroscopicallydetected damage was noted on the RV free wall andthe ventricular septum as a result of the device actua-tion (Fig. 7d), histological examination followinglonger periods of actuation will be required. The im-pact of the device placement and actuation on the leftventricular function will also need to be systematicallystudied. Furthermore, material design optimization forblood compatibility would be required prior to clinicaluse. All these considered, the proposed device wassuccessfully designed for echocardiography guidedimplantation and tested in vitro and in vivo. It repre-sents a promising approach for the support of thefailing RV. It may be especially well situated for therapid establishment of short-term pulsatile RV ejectionaugmentation.

ELECTRONIC SUPPLEMENTARY MATERIAL

The online version of this article (doi:10.1007/s10439-017-1855-z) contains supplementarymaterial, which is available to authorized users.

ACKNOWLEDGMENTS

This work was supported in part by the NIH/NHLBI NCAI Boston Biomedical Innovation CenterPilot Grant (NVV), the DoD CDMRP DiscoveryAward W81XWH-15-1-0248 (NVV), the Wyss Insti-tute for Biologically Inspired Engineering and theHarvard Paulson School of Engineering and AppliedSciences.

REFERENCES

1Anderson, M. B., J. Goldstein, C. Milano, L. D. Morris, R.L. Kormos, J. Bhama, N. K. Kapur, A. Bansal, J. Garcia,J. N. Baker, S. Silvestry, W. L. Holman, P. S. Douglas, andW. O’Neill. Benefits of a novel percutaneous ventricularassist device for right heart failure: the prospective RE-COVER RIGHT study of the Impella RP device. J. HeartLung Transplant. 34:1549–1560, 2015.2Anderson, M. B., and M. O’Brien. Use of the impella 2.5microaxial pump for right ventricular support after inser-tion of heartmate II left ventricular assist device. Ann.Thorac. Surg. 95:e109–e110, 2013.3Anderson, R. H., J. Yanni, M. R. Boyett, N. J. Chandler,and H. Dobrzynski. The anatomy of the cardiac conduc-tion system. Clin. Anat. 22:99–113, 2009.

4Auger, D. A., X. Zhong, and F. H. Epstein. Mapping rightventricular myocardial mechanics using 3D cine DENSEcardiovascular magnetic resonance. J. Cardiovasc. Magn.Reson. 14:1, 2012.5Bramos, D., N. Tsirikos, G. Kottis, C. Pamboucas, V.Kostopoulou, C. Trika, I. Ikonomidis, and S. Toumanidis.The acute effect of an echo-contrast agent on right ven-tricular dimensions and contractility in pigs. J. Cardiovasc.Pharmacol. 51:86–91, 2008.6Buckberg, G. D., and Restore Group. The ventricularseptum: the lion of right ventricular function, and its im-pact on right ventricular restoration. Eur. J. Cardio-Tho-rac. Surg. 29:S272–S278, 2006.7Chou, C. P., and B. Hannaford. Static and dynamiccharacteristics of McKibben pneumatic artificial muscles.IEEE. Int. Conf. Robot. Autom. 1:281–286, 1994.8Chou, C. P., and B. Hannaford. Measurement and mod-eling of McKibben pneumatic artificial muscles. IEEETrans. Robot. Autom. 12:90–102, 1996.9Geva, T., A. J. Powell, E. C. Crawford, T. Chung, and S.D. Colan. Evaluation of regional differences in right ven-tricular systolic function by acoustic quantificationechocardiography and cine magnetic resonance imaging.Circulation 98:339–345, 1998.

10Haddad, F., R. Doyle, D. J. Murphy, and S. A. Hunt.Right ventricular function in cardiovascular disease, partII: pathophysiology, clinical importance, and managementof right ventricular failure. Circulation 117:1717–1731,2008.

11Hsu, D. T., and G. D. Pearson. Heart failure in childrenpart I: history, etiology, and pathophysiology. Circ HeartFail. 2:63–70, 2009.

12Hsu, P., J. Parker, C. Egger, R. Autschbach, T. Schmitz-Rode, and U. Steinseifer. Mechanical circulatory supportfor right heart failure: current technology and future out-look. Artif. Organs 36:332–347, 2012.

13James, T. N., and G. E. Burch. Blood supply of the humaninterventricular septum. Circulation 17:391–396, 1958.

14Lai, W. W., K. Gauvreau, E. S. Rivera, and S. Saleeb.Accuracy of guideline recommendations for two-dimen-sional quantification of the right ventricle by echocardio-graphy. Int. J. Cardiovasc. Imaging 24:691–698, 2008.

15Larose, E., P. Ganz, H. G. Reynolds, S. Dorbala, M. F. DiCarli, K. A. Brown, and R. Y. Kwong. Right ventriculardysfunction assessed by cardiovascular magnetic resonanceimaging predicts poor prognosis late after myocardialinfarction. J. Am. Coll. Cardiol. 49:855–862, 2007.

16Leclair J., M. Doumit, and G. McAllister. Analyticalstiffness modeling and experimental validation for apneumatic artificial muscle. ASME, 2014.

17Markel, T. A., G. M. Wairiuko, T. Lahm, P. R. Crisos-tomo, M. Wang, C. M. Herring, and D. R. Meldrum. Theright heart and its distinct mechanisms of development,function, and failure. J. Surg. Res. 146:304–313, 2008.

18Meyer, P., G. S. Filippatos, M. I. Ahmed, A. E. Iskandrian,V. Bittner, G. J. Perry, M. White, I. B. Aban, M. Mujib, L.J. Dell’Italia, and A. Ahmed. Effects of right ventricularejection fraction on outcomes in chronic systolic heartfailure. Circulation 121:252–258, 2010.

19Obiajulu S. C., and E. T. Roche. Soft pneumatic artificialmuscles with low threshold pressures for a cardiac com-pression device. ASME, 2013.

20Park, E., N. Mehandru, and T. L. Beltran. An intraven-tricular soft robotic pulsatile assist device for right ven-tricular heart failure. J. Med. Dev. 8:020908, 2014.

An Intracardiac Soft Robotic Device for Augmentation

21Roche E. T., and M. A. Horvath. Design and fabrication ofa soft robotic direct cardiac compression device. ASME,2015.

22Roche, E. T., M. A. Horvath, I. Wamala, A. Alazmani, S.-E. Song, W. Whyte, Z. Machaidze, C. J. Payne, J. C.Weaver, G. Fishbein, J. Kuebler, N. V. Vasilyev, D. J.Mooney, F. A. Pigula, and C. J. Walsh. Soft robotic sleevesupports heart function. Sci. Transl. Med. 9:3925, 2017.

23Roche, E. T., R. Wohlfarth, J. T. B. Overvelde, N. V.Vasilyev, F. A. Pigula, D. J. Mooney, K. Bertoldi, and C. J.Walsh. A bioinspired soft actuated material. Adv. Mater.26:1200–1206, 2014.

24Stone, G. W., E. M. Ohman, and M. F. Miller. Contem-porary utilization and outcomes of intra-aortic ballooncounterpulsation in acute myocardial infarction: thebenchmark registry. J. Am. Coll. Cardiol. 41:1940–1945,2003.

25Vasilyev, N. V., J. F. Martinez, and F. P. Freudenthal.Three-dimensional echo and videocardioscopy-guided at-

rial septal defect closure. Ann. Thorac. Surg. 82:1322–1326,2006.

26van de Veerdonk, M. C., T. Kind, J. T. Marcus, G.-J.Mauritz, M. W. Heymans, H.-J. Bogaard, A. Boonstra, K.M. J. Marques, N. Westerhof, and A. Vonk-Noordegraaf.Progressive right ventricular dysfunction in patients withpulmonary arterial hypertension responding to therapy. J.Am. Coll. Cardiol. 58:2511–2519, 2011.

27Voelkel N. F., R. A. Quaife, L. A. Leinwand, R. J. Barst,M. D. McGoon, D. R. Meldrum, J. Dupuis, C. S. Long, L.J. Rubin, F. W. Smart, Y. J. Suzuki, M. Gladwin, E. M.Denholm, D. B. Gail, and L. National heart and bloodinstitute working group on cellular and molecular mecha-nisms of right heart failure. Right ventricular function andfailure: report of a National Heart, Lung, and BloodInstitute working group on cellular and molecular mecha-nisms of right heart failure, 2006.

HORVATH et al.